One tactic that comes to mind for managing online connections is the automated equivalent of an 18th/19th century butler, who mediated social interaction for the wealthy at a time when the intrusive, in-person visit was a primary method for making social contact.
The butler has broad and nuanced knowledged of the circumstances in which the Lady is to be acknowledged to be IN.
Today’s online presence indicators are flat; they tell everyone the same message; that one is working, or eating ice cream in front of the television, or AWAY.
A butler would understand whether one is working or not, and would put through different connections at different hours.
A butler would understand the understand the nuances of one’s social circle, and admit some people automatically, allowing others to wait, and requiring still others to leave a message.
We also need the 18th/19th century interface to the Butler, the calling card, which conveys by its printed message and accompanying whether the caller is a longlost relative whom the butler may not remember; a recommendation from a reputable source; a specific message about the urgency of the visit.
Along with butlers and calling cards would come social norms for interpreting the signals — when a calling card is a polite formality; when to interpret the declining of a visit as a crippling snub and when as scheduling circumstance. Even with a butler, there will be mistimings and misunderstandings, yielding to new materials for comedy and drama.
Adina,
Reading your post I realized that what we have in social triage (see http://www.lifewithalacrity.com/2005/02/dunbar_triage_t.html)
is an information retrieval problem over your personal space of social resources (friends, contacts, associates, whatever – they’re people not text). The goal is to be at least Pareto Optimal in achieving both precision and recall in your social networking tools.
A social network must ideally retrieve all relevant social connections while excluding precisely those that are irrelevant. Its a tradeoff they can’t make work well enough in information retrieval without human intervention hence Google’s success with PageRank which relies on webmaster’s recommendations with links. I bet the social networks that work best will elicit some kind of PageRank-link social recommendation to unflatten the spacial of possible social connections.
Best,
Richard
Whoa. I like that, even if I can only half get it. I have just learned what precision and recall mean in IR terms, but Googling for Pareto Optimal I don’t see how to map that concept onto precision and recall.
Then there is the leap of declaring the nuanced management of attention and social relationships an IR problem, as opposed to a problem in some other constellation of algorithms. And the further leap of the analogy with PageRank. I can see how PageRank could apply if, say, I wanted to find a friend of a friend to recommend a restaurant or review my resume, but I don’t see how it applies to the problem of my cell phone knowing whether to ring or go straight to voicemail.
I hope Richard will stop by and explain further, or that Adina got it and can do so for him.
The “relevance” associated with social attention seems different from the relevance associated with information retrieval.
IR relevance is about the relatedness of a word to a topic. A search for “apple” is about a fruit, or a make of computer. There are various ways to infer from a text whether it’s about one or the other.
Social relevance is about the closeness of the person with respect to some domain: people I trust for computer advice, musical taste, personal discretion. This is harder (or impossible) to measure by word search.
A search engine can tell on its own whether an acquantaince has similar musical taste, or whether they have an interest in social software. But the search engine could not tell whether I consider a person to be reliable or ethical.
Social proximity is a good proxy. I’m more likely to trust friends of friends. But the information that represents social trust is not very accessible to search engines.
Agreed, Adina. And then your Butler metaphor brings in an equally complicated problem, namely how to automate the matter of interruption management. Social proximity doesn’t map easily into a decision about whether I want to be interrupted for some synchronous communication (in other words, whether the phone should ring or cut to voicemail). For one thing, what I’m doing at the moment (working, sleeping, showering, eating ice cream in front of the TV) figures into the decision. And even if I sat down with a big sheet of gridded paper and put everyone I know on one axis and all my daily activities on another, I couldn’t come up with an adequate set of fixed rules. That grid would leave out whether I might be expecting a call from a particular person, for instance. And let’s not forget my mood! The classic (and perhaps mythical) butler would adapt the rules on the fly, often without my having to say a word.
These sorts of nuances are why I’ve never believed in the automated scheduling agents that have long been touted by everyone from Nicholas Negroponte to the current proponents of the Semantic Web. Even the relatively simpler problem of scheduling a dentist’s appointment has too many dependencies for me to simply point a souped-up iCal at the dentist’s Semantic Web-enhanced site and let the two of them negotiate a decision.